Pinned Tweet
- CPU: AMD Threadripper PRO 5955WX (16 cores / 32 )
- Mobo: ASUS Pro WS WRX80E-SAGE SE WIFI
- RAM: 512GB DDR4-2666 ECC
- GPUs:
- 3x NVIDIA RTX 5090 32GB
- 1x NVIDIA RTX 3090 24GB
- 120GB VRAM total
- Storage:
- Samsung 990 EVO Plus 4TB
- SK hynix 512GB
- PSU: 2500W
1
249
so, it has come to my understanding, that the game most of you cucks have been waiting for, has come out, and it's a literal cuck simulator game?
7
What if I told u that israel is behind those cyber attacks, because Trump doesnt want to continue the Palestinian genocide?
33
I tied Fable 5.2 today. It didn't assume I'm building a nuclear bomb that will release some toxic bio cloud while I'm building a trading API.
Good times
82
I built a CVE Patcher for upstream CVEs without an official patch due to how many there are 😅
github.com/ashhart/CVE-patch
curl -fsSL raw.githubusercontent.com/as… | sh -s -- --patch CVE-2026-98432
CryptoScene retweeted
🧲 Community torrents · 28 Oct 2026
60 model torrents added:
• Comfy-Org/MOSS-Image
• huihui-ai/Huihui-Qwen3.8-Flash-Next-abliterated-GGUF
• Comfy-Org/Z-Image
• Qwen/Qwen3.8-Flash-Next-FP8
• Qwen/Qwen3-30B-A3B-Thinking-2507
• mistralai/Mixtral-8x22B-Instruct-v0.1
• hum-ma/Wan2.2-TI2V-5B-Turbo-GGUF
• llmfan46/Qwen3.5-9B-Ultra-Uncensored-Heretic-GGUF
• mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit
• Qwen/Qwen2-57B-A14B
• JunhaoZhuang/FlashVSR
• audio-cpp/Ming-3.0-Audio-Tiny-GGUF
• microsoft/llava-med-v1.5-mistral-7b
• pottokao/MOSS-Image-Text-Encoder-Heretic-DHQ-GGUF
• huihui-ai/Huihui-Ornith-1.5-35B-A3B-abliterated
• OS-Software/gemma-4-E4B-it-uncensored-heretic-GGUF
• ReadyArt/gemma-4-31B-it-scotoma-2
• RBDash-Team/Xing-VL2.0-Decision
• bottlecapai/ThinkingCap-Qwen3.8-Flash-Next-GGUF
• fastino/Fastino-Nemotron-3.5-Lightning-Finance
• OpenMOSS-Team/MOSS-Audio-8B-Instruct
• Naphula/Cyberdonia-24B-v8
• inclusionAI/LLaDA2.2-flash
• apetersson/DeepSeek-V4.1-Flash-1017-Abliterated-MLX
• darkc0de/Muse-Glimmer-30B-heretic
• KRAFTON/Raon-1-mini-preview-dflash2
• securepeak/DeepSeek-V4.1-Flash-Abliterated
• Jackrong/Kestrel-4B-Coder-GGUF
• Agnes-AI/Agnes-3.0-Qwen
• FINAL-Bench/POCKET-EN-GGUF
• OpenLLM-France/Luciole-8B-Base
• OpenLLM-France/Luciole-23B-Audio-1.0
• ZeroWw/Meta-Llama-3.1-8B-Claude-39fail-3000total-GGUF
• tencent/Hy-SimulMT-30B-A3B-FP8
• ML-Intern-lab/agate-preview-002-4step
• tencent/Hy-SimulMT-30B-A3B-Interleave
• OpenLLM-France/Luciole-23B-Thinking-1.1
• OpenLLM-France/Luciole-1B-Audio-1.0
• VAGOsolutions/SauerkrautLM-MoEdernColBERT-1.5M
• vllm-sr/vela3-fast
• mrlarp/larrylarp3-btc-avg9-15m
• mrlarp/larrylarp3-btc-avg9-1h
• mrlarp/larrylarp3-btc-avg9-5m
• mrlarp/larrylarp3-btc-avg9-4h
• jesdga/Qwen3.8-Flash-Next-Basalt
• ShyamVarahagiri/MachineTranslation
• cheppard03/gemma-4-E4B-it-OBLITERATED
• basecompute/gemma-4-E4B-it-BaseQuant-V1
• bendernina/gemma-4-E4B-it-OBLITERATED
• Edge0/Audio8-TTS-Realistic-0.6b
• + 10 more
Browse new magnets ↓
Been running Basalt vs Strata on the same Qwen3.8-Flash-Next pack, same boxes. 2x5090: 221-328 t/s vs ~200.
One card as a second tenant, no problem. And context is elastic, so 256K came out of a config flag, 240K prompt, needle still recalled.
Same weights, fewer cards, faster
1
108











